VeriMask: Facilitating Decontamination of N95 Masks in the COVID-19 Pandemic: Challenges, Lessons Learned, and Safeguarding the Future

Page created by Norman Sanders
 
CONTINUE READING
VeriMask: Facilitating Decontamination of N95 Masks in the COVID-19 Pandemic: Challenges, Lessons Learned, and Safeguarding the Future
VeriMask: Facilitating Decontamination of N95 Masks in the
COVID-19 Pandemic: Challenges, Lessons Learned, and
Safeguarding the Future

YAN LONG, EECS Department, University of Michigan, USA
ALEXANDER CURTISS, ECE Department, Northwestern University, USA
SARA RAMPAZZI, CISE Department, University of Florida, USA
JOSIAH HESTER, ECE & CS Department, Northwestern University, USA
KEVIN FU, EECS Department, University of Michigan, USA
The US CDC has recognized moist-heat as one of the most effective and accessible methods of decontaminating N95 masks
for reuse in response to the persistent N95 mask shortages caused by the COVID-19 pandemic. However, it is challenging
to reliably deploy this technique in healthcare settings due to a lack of smart technologies capable of ensuring proper
decontamination conditions of hundreds of masks simultaneously. To tackle these challenges, we developed an open-source
wireless sensor platform—VeriMask1 —that facilitates per-mask verification of the moist-heat decontamination process.
VeriMask is capable of monitoring hundreds of masks simultaneously in commercially available heating systems and provides
a novel throughput-maximization functionality to help operators optimize the decontamination settings. We evaluate VeriMask
in laboratory and real-scenario clinical settings and find that it effectively detects decontamination failures and operator errors
in multiple settings and increases the mask decontamination throughput. Our easy-to-use, low-power, low-cost, scalable
platform integrates with existing hospital protocols and equipment, and can be broadly deployed in under-resourced facilities
to protect front-line healthcare workers by lowering their risk of infection from reused N95 masks. We also memorialize the
design challenges, guidelines, and lessons learned from developing and deploying VeriMask during the COVID-19 Pandemic.
Our hope is that by reflecting and reporting on this design experience, technologists and front-line health workers will be
better prepared to collaborate for future pandemics, regarding mask decontamination, but also other forms of crisis tech.
CCS Concepts: • Computer systems organization → Embedded systems.
Additional Key Words and Phrases: Wireless Sensor, COVID-19, N95 Masks Decontamination
ACM Reference Format:
Yan Long, Alexander Curtiss, Sara Rampazzi, Josiah Hester, and Kevin Fu. 2021. VeriMask: Facilitating Decontamination of
N95 Masks in the COVID-19 Pandemic: Challenges, Lessons Learned, and Safeguarding the Future. Proc. ACM Interact. Mob.
Wearable Ubiquitous Technol. 5, 3, Article 119 (September 2021), 29 pages. https://doi.org/10.1145/3478105
1 We release our open-source design at https://github.com/longyan97/VeriMask_Designs

Authors’ addresses: Yan Long, EECS Department, University of Michigan, Ann Arbor, Michigan, USA, yanlong@umich.edu; Alexander
Curtiss, ECE Department, Northwestern University, Evanston, Illinois, USA, alexandercurtiss2025@u.northwestern.edu; Sara Rampazzi,
CISE Department, University of Florida, Gainesville, Florida, USA, srampazzi@ufl.edu; Josiah Hester, ECE & CS Department, Northwestern
University, Evanston, Illinois, USA, josiah@northwestern.edu; Kevin Fu, EECS Department, University of Michigan, Ann Arbor, Michigan,
USA, kevinfu@umich.edu.

Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that
copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page.
Copyrights for components of this work owned by others than the author(s) must be honored. Abstracting with credit is permitted. To copy
otherwise, or republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee. Request permissions from
permissions@acm.org.
© 2021 Copyright held by the owner/author(s). Publication rights licensed to ACM.
2474-9567/2021/9-ART119 $15.00
https://doi.org/10.1145/3478105

 Proc. ACM Interact. Mob. Wearable Ubiquitous Technol., Vol. 5, No. 3, Article 119. Publication date: September 2021. 119
VeriMask: Facilitating Decontamination of N95 Masks in the COVID-19 Pandemic: Challenges, Lessons Learned, and Safeguarding the Future
119:2 • Yan Long, Alexander Curtiss, Sara Rampazzi, Josiah Hester, and Kevin Fu

1 INTRODUCTION
N95 mask shortages have been a persistent and enduring problem in almost every significant pandemic. The SARS
outbreaks in 2003, H1N1 influenza in 2009 [60, 100], and the COVID-19 pandemic all had severe N95 shortages.
During each pandemic, the abrupt and fierce increase in the demand for N95 masks quickly depleted the supply
chain, forcing the public and healthcare workers to reuse their masks. The ongoing COVID-19 pandemic has
seen particularly serious and prolonged N95 mask shortages worldwide. For instance, a survey in the US from
early 2020 [29] shows that over 50% of 21,000 nurses surveyed were required to reuse N95 masks for at least 5
days on average. The mask shortage has been plaguing the US even until early 2021 [16, 35] and has continued to
severely impact India [11, 38] and other countries [34, 37] suffering from the new wave of COVID-19 variants.
The reuse of N95 masks poses severe risk of infection because of the potential pathogenic agents present on the
masks, especially for front-line healthcare workers using the same disposable mask over multiple days when
treating patients [31].
 While these cycles of N95 shortages in recent pandemics highlight the poor preparedness of personal protection
equipment (PPE) supply chain management [60, 100], it also raises a question for researchers and system
developers: What can be done in a pandemic to quickly respond to N95 mask shortages, and how can researchers
engage with communities to prepare for future shortages in the next pandemic? The medical research community
has proposed several methods for decontaminating single-use N95 masks for safe reuse, such as hydrogen peroxide
dosing, autoclave treatment, gamma and UV-C irradiation, etc [101]. These methods, however, either suffer
from chemical exposure and potential harm to the wearer caused by toxic residuals [30], or require expensive
specialized equipment costing as much as $6 million USD [25].
 In this work, we identify a candidate for safe, low-cost, and efficient N95 mask decontamination, i.e., the
moist-heat decontamination method, and study the critical challenges preventing this decontamination method
from being rapidly, widely, and reliably deployed in healthcare facilities for protecting front-line healthcare
workers. Moist-heat decontamination has been recognized by the US CDC as one of the most effective and
easy-to-operate decontamination methods for the SARS-CoV-2 virus [18], and has the potential to be deployed
using commercial heating devices such as ovens and warming cabinets that are already widely available in
hospitals [17]. Such deployment, however, is challenging because of:
 (1) the risks of decontamination failures caused by non-uniform heating and humidity fluctuations in the
 commercial heating systems [45, 78, 80],
 (2) unpredictable operating errors, due to human error stemming from complex protocols and exhaustion,
 (3) the large burden and difficulty of manual verification of each mask’s decontamination status,
 (4) and the impracticality of employing closed-loop control to achieve maximal decontamination throughput.
 To address these challenges, we developed VeriMask, an open-source2 wireless sensor platform that enables
rapid, reliable, and cost-efficient moist-heat decontamination in healthcare facilities. VeriMask incorporates
low-cost BLE-enabled sensor nodes in a dense sensing topology that continuously monitors the temperature
and relative humidity of each mask and an Android application that is capable of automatically verifying the
decontamination status of hundreds of masks simultaneously. Based on decontamination profiling, the Android
application is also able to determine and suggest the optimal decontamination settings and help the operators to
maximize the decontamination throughput. In order to be applicable in a wide range of decontamination scenarios
and heating devices, VeriMask is designed to be flexible in terms of configurable decontamination parameters.
VeriMask integrates with existing clinical workflows, and the wireless feature ensures scalability of VeriMask
so that it can be rapidly deployed in clinical settings at large scales. The easy-to-use VeriMask App provides
decontamination information of each individual mask and guarantees adequate decontamination before the N95
masks can be reused. The low-power, high temperature resistant clinical-use VeriMask sensor nodes with a size
2 We release our open-hardware and open-source design at https://github.com/longyan97/VeriMask_Designs.

Proc. ACM Interact. Mob. Wearable Ubiquitous Technol., Vol. 5, No. 3, Article 119. Publication date: September 2021.
VeriMask: Facilitating Decontamination of N95 Masks in the COVID-19 Pandemic: Challenges, Lessons Learned, and Safeguarding the Future
Facilitating Decontamination of N95 Masks • 119:3

of 3.5 x 3.8 cm (1.4 x 1.5 inches) can work continuously for over 1000 decontamination cycles without change
of battery and only cost $15.66 each at a quantity of 1000. We evaluated VeriMask in both a laboratory setting
and a clinical setting. Results show that VeriMask is able to detect subtle decontamination failures that might go
undetected without accurate verification and enhance efficiency by increasing decontamination throughput.
 VeriMask creates a new system using well-known and verified components, which enables it to have a
real impact on real-world clinics. VeriMask is the first realization of an open-source sensor system design for
disposable mask decontamination, which can play critical roles in mitigating mask shortage emergency caused
by current and future pandemics. To ensure that the challenges, lessons, and design outcomes from VeriMask’s
design and methodology can transfer to future open-source design for emergency response, we constructed
a design framework from VeriMask’s design criteria by relating to researches and works of Free and Open
Source Hardware (FOSH) and proposed general design goals that can be widely applied. Finally, we glean lessons
from our experience with building and deploying VeriMask, and the experience of our clinical and near-clinical
collaborators and partners. We envision that our highly adaptable open-source VeriMask platform and our
experience can greatly accelerate the design and deployment process of mask decontamination sensor systems
and safeguard society when the next pandemic hits. We hope that the general lessons learned from mobile
computing researchers responding to a societal-scale emergency will be memorialized and aid future researches
hoping to respond to future crises. We summarize our main contributions as follows:
 (1) We designed and built the VeriMask platform specifically conceived to verify moist-heat decontamination
 processes of N95 masks in a semi-automated, human-in-the-loop, reliable way. VeriMask is flexible, scalable,
 low-power, and easy-to-use for non-specialized operators, which enables rapid deployment in under-
 resourced clinical settings to protect front-line health workers. VeriMask is self-healing and adjusts to
 diverse types of heating devices using a profiling function and operator feedback.
 (2) We evaluate VeriMask in laboratory and clinical settings and demonstrate its effectiveness in detecting
 unsuccessful decontamination that might remain undetected using conventional approaches. We also
 integrate VeriMask into existing emergency operating protocols to enable fast, low-cost, and reliable
 deployment of moist-heat decontamination processes with commercial heating devices in hospitals.
 (3) We formulate VeriMask’s design criteria in a transferable way and elaborate our design experience, con-
 siderations, and lessons learned to help future designers quickly respond to mask shortage (and other
 large-scale societal) emergencies with open-source design.

2 BACKGROUND & RELATED WORK
2.1 N95 masks decontamination methods
Disposable N95 Filtering Facepiece Respirators (FFRs) commonly referred to as N95 masks or N95 respirators,
are worn by healthcare workers for self-protection. These masks are composed of multiple layers including
an electrostatic filter, depending on the model. With more than 95% filtration efficiency, they are designed to
prevent the wearer from inhaling small airborne particles by capturing them via mechanical and electrostatic
forces [1, 20]. N95 respirators are "single-use," disposable devices that should not be shared or reused.
 The US Centers for Disease Control and Prevention (CDC) indicates that N95 masks can be used for up to 8
hours (including between patients), then the device should be properly discarded and replaced [24]. Over time,
the warm and humid environment caused by breathing can accelerate the spread of the captured microorganisms
to the inner layers of the mask, posing a risk of contaminant exposure for the wearer [99].
 The shortage of N95 masks during the SARS-CoV-2 pandemic has prompted the consideration of reuse of these
devices after a decontamination process to extend limited stocks [18]. The US Food and Drug Administration
(FDA) guidance identifies as a target for adequate decontamination a 3-log level, or more, bioburden reduction.
In other words the process should reach a viral load reduction of at least a factor of 1000 [26]. In addition, an

 Proc. ACM Interact. Mob. Wearable Ubiquitous Technol., Vol. 5, No. 3, Article 119. Publication date: September 2021.
VeriMask: Facilitating Decontamination of N95 Masks in the COVID-19 Pandemic: Challenges, Lessons Learned, and Safeguarding the Future
119:4 • Yan Long, Alexander Curtiss, Sara Rampazzi, Josiah Hester, and Kevin Fu

 3KDVH 3KDVH 3KDVH

 0DVN%DWFK 'HFRQWDPLQDWLRQ 0DVN8QORDG 
&RQWDPLQDWHG 3UHSDUDWLRQ 6HW8S
 /RDG 6HOHFWLRQ
1PDVNV 'HFRQWDPLQDWHG
 10DVNV
 +HDWLQJ &RQYHFWLYH
 0DVNEDWFK
 V\VWHP +HDWLQJ6\VWHP 9HUL0DVN
 5HDOWLPH
 XQORDG
 VHWWLQJ 1RGH1RGH

 PRQLWRULQJ
 GHFRQ 1RGH1RGH

 FRQGLWLRQ 9HUL0DVN 'LVFDUGHG
 9HUL0DVN 6LQJOHPDVN
 1RGH1RGH
 10DVNV
 YHULILFDWLRQ
 VHWWLQJ GHFRQ
 FKHFN

 10DVNVWR
 UHSURFHVV

Fig. 1. Three phases of a typical moist-heat decontamination cycle integrated with VeriMask. Phase 1: Each contaminated
mask is put into a rigid container with a VeriMask device and water soaked paper towel to create adequate humidity. Each
sealed container is then transferred into the heating system. Phase 2: The Android app monitors the decontamination
process in real time to detect potential anomalies (e.g. a container temperature dropped below threshold). Phase 3: After
the decontamination time the operator check the type of failure flagged by VeriMask. If the masks that have experienced
decontamination failure can be re-processed, they will be integrated in the next decontamination cycle, otherwise they will
be discarded.

effective N95 mask decontamination should maintain the fit, sealing capacity, and the filtration performance of
the mask. Finally, the process should not damage the mask’s structural integrity, material, and should present no
residual chemical hazard for the wearer due to the treatment (e.g. skin irritation, respiratory distress) [18].
 Different approaches has been investigated to decontaminate disposable N95 masks to mitigate the shortage due
to the massive demand, as shown in Table 1. In particular, the US National Institute for Occupational Safety and
Health (NIOSH) [18] and teams of researchers [39, 57, 90, 93] have identified Ultraviolet Germicidal Irradiation
(UVGI), hydrogen peroxide, and moist-heat (MH) as the most promising methods to decontaminate N95 masks
against the virus SARS-CoV-2 without affecting the filtration capacity, fit, and seal.
 While Hydrogen peroxide-based methods are widely used in hospitals for inactivating highly resistant
pathogens [101], these processes require expensive and specialized equipment [25] and careful control of humidity
level, gas saturation, concentration, and duration of exposure. The entire process has a duration up to 8 hours,
and only trained personnel should operate the equipment because errors in the dosing protocols could result in
decontamination failure and even explosion hazards [90]. In addition, hydrogen peroxide gas is a corrosive irritant
that can interact with N95 mask to form toxic residues to wearers if not properly removed after treatment [15].
UVGI-based methods depends critically on the UV-C source wavelength (200-280 nm with peak efficacy at ~254
nm) and UV-C dose (≥1.0 / 2 ) applied to each mask surface [57]. Previous studies [61, 73, 103] have found
that (i) the layer structure of certain N95 mask models and the presence of shadows can compromise the efficacy
of the treatment because the mask surfaces may not receive the full UV-C dose, (ii) the mask straps required a
further decontamination procedure because of residuals, (iii) direct exposure to UV-C light is harmful and UV
wavelengths in the range of 175–210 nm can generate ozone which is hazardous to human health.
 In contrast to hydrogen peroxide and UVGI methods, moist-heat treatments have minimal duration (typically
30-60 minute), do not require specialized and expensive equipment or a separate decontamination procedure
for the straps, and do not require an off-gas time or a further treatment for removing dangerous chemical
residues [39, 48, 79, 92]. For these reasons, MH methods can be suitable for wide and fast deployment in hospitals,
laboratories, and non-specialized healthcare facilities that are generally already equipped with commercial

Proc. ACM Interact. Mob. Wearable Ubiquitous Technol., Vol. 5, No. 3, Article 119. Publication date: September 2021.
VeriMask: Facilitating Decontamination of N95 Masks in the COVID-19 Pandemic: Challenges, Lessons Learned, and Safeguarding the Future
Facilitating Decontamination of N95 Masks • 119:5

Table 1. Comparison between different N95 masks decontamination methods for on-site deployment. Moist-heat decontami-
nation does not require expensive equipment or off-gas time to remove harmful gas or chemical residues.
 Decontamination Chemical Operator Process Costs** Throughput Max Reuse Literature
 Method Residue Hazard* Time (N95/Cycle) Cycles sources
 VHP Yes Chemical 4-8h $$$ ≤ 160 4+ [12, 19]
 HPGP Yes Chemical 24min-1h $$$ ≤ 10 2+ [13, 19]
 UVGI No Ozone Exposure 10min*** $$ ≤30++ 5 [10, 32]
 Autoclave Steam No No 30min-1h $$$$ 12-20 4+ [28, 44]
 Moist Heat No No 30min-1h $ ≤ 100++ 5 [19, 39]
 *Assuming standard protection procedures are followed (e.g. wearing mask, gloves, long-sleeved gown, eye protection).
 **Considering startup and recurring costs: $$$$ = >$30,000, $$$ = $5,000 - $30,000, $$ = $1,000 - $5,000, $ =
VeriMask: Facilitating Decontamination of N95 Masks in the COVID-19 Pandemic: Challenges, Lessons Learned, and Safeguarding the Future
119:6 • Yan Long, Alexander Curtiss, Sara Rampazzi, Josiah Hester, and Kevin Fu

3 OPEN SOURCE DESIGN FOR EMERGENCY DECONTAMINATION
The failure of traditional production methods to meet the massive demand for PPE during the COVID-19 pandemic
has led to rapid response from the scientific community on developing new technologies and tools for emergency
use. In particular, Free and Open Source Hardware (FOSH) has been recognized as beneficial in emergency
situations due to its high accessibility and the collective contributions from worldwide expertise which enables
fast development [75, 81]. VeriMask is the first realization of open source design for safe verification of moist-heat
decontamination that is fully integrated with clinical workflows, and was broadly shared with a volunteer
collective [89](registered 501(c)(3) non-profit) of scientists, engineers, clinicians, and students from universities
and health systems across the world and professionals in the private sector. In this section, we describe the
design criteria, challenges, and specifications we formulated developing VeriMask based on our experience and
interaction with stakeholders in this collective.

3.1 Foundational Design Criteria & Guidelines
Leveraging the design experience and critical implementation factors highlighted by previous FOSH works and
open source medical hardware design [36, 47, 62, 76, 91, 96], we have formulated the following 3 foundational
criteria and corresponding guidelines for designing VeriMask. These criterion and guidelines are, in many ways,
generally applicable to emergency response hardware design.
Criterion 1: Fast design & deployment. The emergency caused by shortages poses severe risks of infection
to health workers due to the prolonged use and reuse of PPE while treating sick patients. A timely design and
deployment for protection thus become crucial [53, 84]. Designing simple and highly debuggable open hardware
devices significantly accelerates the implementation, deployment, and maintenance processes [47, 76], and helps
build a trustworthy relationship between designers and healthcare facilities [47]. In addition, utilizing common
off-the-shelf (CoTS) components built and tested by trustworthy manufacturers and available in bulk facilitates
implementation and enhances hardware reliability by avoiding long-time testing, calibration, and availability
problems due to custom hardware. Unlike devices with novel sensing functionality or materials, it is imperative
that emergency response open source design use "boring" but reliable devices to ensure real-world use.
Criterion 2: Generality-oriented robust design. The desired generality is two-fold. First, the function-specific
design should be made applicable to a large scope of deployment scenarios, e.g., using various types of heating
devices. The actual use case of the design in hospitals is usually unpredictable because of the variance in the
existing equipment that different healthcare facilities already posses. In addition, the designers often do not have
direct access to the deployment environment due to healthcare regulations [53, 54], which further undermines
the feasibility of a dedicated design. It is thus crucial in the design process to assume a weak dependence on the
deployment environment and relevant equipment so that it can be easily and widely adopted by different entities
for the target purpose. Second, the building blocks of the open source design should have general and common
enough functionality plus interface. The design should be able to integrate well with existing infrastructures. For
instance, employing popular communication schemes such as WiFi, Bluetooth, and USB not only reduces the
cost of the system by eliminating DIY communication modules, but also enables easier development, debugging,
replacement, and use. Ensuring such general and common interface enables the designers to easily replace
the components in case of a local supply-chain shortage. Generally speaking, ensuring a generality-oriented
design also helps make the open source design transferable and reusable for future systems, and makes parts
obsolescence (common in the electronics industry) less likely.
Criterion 3: Priority-centered design. The most important criterion for making such healthcare-related open
source design is to enforce a priority-centered principle in the specification formulation process. To achieve
that, we propose 3 categories of design goals ordered according to their priorities. 1) Safety. Safety requirements

Proc. ACM Interact. Mob. Wearable Ubiquitous Technol., Vol. 5, No. 3, Article 119. Publication date: September 2021.
VeriMask: Facilitating Decontamination of N95 Masks in the COVID-19 Pandemic: Challenges, Lessons Learned, and Safeguarding the Future
Facilitating Decontamination of N95 Masks • 119:7

are those that can directly endanger the operators and/or intended users if not met. For example, being able to
accurately and reliably determine whether each mask has been successfully decontaminated or not is a safety
requirement for our mask decontamination system design. Safety requirements are the most important and must
be satisfied before other specifications can be considered [23, 36, 91]. To ensure safety, the designers should
fully understand the standards and requirements in government regulations and academic publications, choose
the sensor monitoring scheme based on the worst-case scenarios, and carefully evaluate the information loss
and consequences in sensor data transmissions. 2) Effectiveness. We define effectiveness as a combination of
the degree of achievement for criterion 1 and 2, as well as other performance factors such as designed system’s
throughput, cost, battery life, and size (in the case where portable devices are desired). Considering the large
number of form factors, the designers should develop rubrics for prioritizing the factors and understanding
performance. Besides ensuring safety and fast development of VeriMask, we determine scalability, low-cost, and
small-size as the top-3 factors to make sure under-resourced hospitals can quickly deploy the platform in large
scale. 3) Usability. Usability includes factors such as reducing the workload of the users, making the design easy
to use and understand for non-expert operators, and making the devices more comfortable to use (especially in
the case of wearable devices). Designers should always design with the clinical operating protocol in mind for
improving the actual usability. To ensure high usability of VeriMask, we target at automating the process of mask
decontamination verification, maximizing the decontamination throughput, and minimizing the workload of
configuring and maintaining the VeriMask platform. In general, these priority mappings hold across emergency
response design, as safety, effectiveness, and usability can serve as a rubric for evaluating possible solutions.

3.2 Challenges of MH Decontamination
The most effective methodology of moist-heat decontamination that has been validated on different N95 mask
models consist of treating the contaminated masks by holding them within a target temperature range 
VeriMask: Facilitating Decontamination of N95 Masks in the COVID-19 Pandemic: Challenges, Lessons Learned, and Safeguarding the Future
119:8 • Yan Long, Alexander Curtiss, Sara Rampazzi, Josiah Hester, and Kevin Fu

have exceeded the maximal allowed number of decontamination cycles. Since the inspection is only based on
visual characteristics, the operator is unable to determine if the decontamination has been unsuccessful (e.g. with
temperature or humidity drop) so the masks are still contaminated. An effective verification tool needs to address
the challenge of enabling the operators to conduct accurate and fast verification by providing reliable feedback.
Non-specialized Commercial Devices and Unpredictable Failures. One major advantage of MH decon-
tamination is its use of commercial heating devices equipped by hospitals which are considered suitable for
emergency decontamination [17]. However, such non-specialized heating devices inherently suffer from the
problem of non-uniform heat distribution as well as complex and uncontrolled heating dynamics [45, 78, 80, 102].
For example, considering the total MH process time as the sum of the transient time and decontamination
time, i.e., = + , we found with a clinical humidity chamber more than 20 minutes of transient time
variation for reaching the decontamination temperature in different zones (Figure 2 (a)), and more than 9 degrees
of difference between the device’s set temperature and the internal temperature of the containers even when the
cabin was pre-heated (Figure 2 (b)). Such heating dynamics are also affected by the load and initial temperature
conditions. In addition, unpredictable failures such as humidity leakage can also happen due to operational errors
during the treatment. These challenges make a static preventive device profiling using conventional thermal
profilers [33] insufficient and call for real-time verification technologies. We discuss these factors in Section 6.
Throughput Optimization. Another challenge of deploying efficient MH decontamination processes is achiev-
ing optimal decontamination throughput, i.e., ensuring the maximum number of successfully decontaminated
masks in given time. Ideally, this could be achieved by a full control of the heating device to dynamically adjust
the temperature and ventilation of the cabin. However, the large variations in heating devices hardware and
software as well as the impracticality of a nurse or healthcare worker modifying/customizing commercial heating
devices make it infeasible to realize such a closed-loop control while maintaining wide applicability, fast &
reproducible design, and low cost. For these reasons, solving the challenge requires developing a reliable method
for predicting the decontamination throughput and providing the user with the optimal settings that achieve
throughput maximization without being specific to a particular heating device.

3.3 MH Decontamination Process Integrated With VeriMask.
We designed a clinical workflow, with input from the volunteer collective, that would be as low burden as possible
using VeriMask. As shown in Figure 1, the typical operating procedures of a MH decontamination cycle with
VeriMask can be broken down to three main phases [17]. During phase 1, the operator preheats the heating device
to the target temperature (~80°C) based on the adopted MH treatment, and configures the VeriMask smartphone
app with the appropriate thresholds for temperature ( ℎ and ), humidity ( ℎ and ), and decontamination time
( ). Each contaminated N95 mask is then put into a separate rigid container (e.g. polypropylene, or oven
safe Pyrex glass) with (a) the VeriMask sensor node attached to the lid to avoid direct contact with the masks,
and (b) a paper towel soaked with 500 L of water (see Figure 4). Then, each container is sealed by properly
closing the lids and transferred into the heating system for the treatment. The duration of this phase can take
approximately 20-30 minutes depending on the mask batch size. Phase 2 consists of waiting for the predetermined
MH process time for one cycle of decontamination. In this delicate phase, VeriMask automatically checks
if the critical process conditions are met and declares decontamination failures if certain anomaly conditions
happen (e.g., the relative humidity level falls below the lower threshold). This phase can take 40 to 60 minute
based on the MH process adopted, masks’ model, and heating device used. Finally, in phase 3 the operator removes
the containers from the heating device and sorts each mask based on the mask status shown on the VeriMask
monitor application. The masks can either be returned safely to the original user, discarded, or reprocessed if
necessary based on the clinical policy. This last phase might have a duration of 5-10 minutes depending on the

Proc. ACM Interact. Mob. Wearable Ubiquitous Technol., Vol. 5, No. 3, Article 119. Publication date: September 2021.
VeriMask: Facilitating Decontamination of N95 Masks in the COVID-19 Pandemic: Challenges, Lessons Learned, and Safeguarding the Future
Facilitating Decontamination of N95 Masks • 119:9

 ts2
 ts1
 24min

 70 Tl 80.4C 80.1C
Temp (°C)

 78.9C
 17°C
 50 75.1C
 76.5C
 Top Shelf 70.6C
 30 Middle Shelf
 Bottom Shelf 75.0C
 10 80.0C 79.5C
 0 10 20 30 40 50 60 70 79.3C

 Time (min)
 (a) (b)

Fig. 2. (a) Temperatures and transient times measured with VeriMask on three different shelves of a clinical humidity
chamber during the heating process (starting from 19°C room temperature). The middle shelf container reaches the target
temperature of 70°C 24 minutes after the containers on the top and bottom shelves. (b) The 10 locations chosen to measure
the non-uniform heat distribution in the pre-heated clinical humidity chamber. The temperature values (in red) show more
than 9 degrees of difference in the internal temperature of the containers even 30 minutes after the cabin reached the target
temperature (82°C) measured by the heating device built-in sensor.

mask batch size. Once each cycle is ended, the sensor nodes can be reset in the app, and then the containers can
be filled again with a new batch of masks to decontaminate in the next cycle.
Profiling cycle. Every cycle monitored via VeriMask can be used as a profiling cycle. When this functionality is
enabled using the application, the operator can input the expected total working time defined as the continuous
decontamination time in a day (e.g. 8 hours). The temperature and humidity information collected during this
profiling cycle (with a default process time of = 50 ) are then used to compute the optimal heating
device temperature and process time for achieving the maximum number of decontaminated masks in the
selected working time. The application also provides information about container locations not suitable for
decontamination so that operators can avoid placing masks in these areas. If the operator confirms the optimization
parameters, the process time for following decontamination cycles are automatically adjusted.

4 VERIMASK DESIGN
Figure 3 shows the architecture of the VeriMask platform. The platform consists of two major components:
wireless sensor nodes and an application running on smartphones. The nodes, each one associated with an N95
mask, periodically send temperature and humidity data to the application which performs real-time monitoring,
verification of the decontamination conditions, and optimization of the overall decontamination process.

4.1 VeriMask Sensor Node
We designed and manufactured the VeriMask sensor node V1 for experimental testing and performance evaluation,
and node V2 for practical clinical usage and deployment. The two versions of sensor nodes are shown in Figure 4.
 The VeriMask sensor nodes have been designed to resist the harsh environmental condition of repeated
heating cycles by selecting components and making the optimal layout to be high temperature resistant. Each
sensor node consists of a Laird653 wireless module with a Nordic nRF52833 SoC [7] that features a maximum
operating temperature of 105°C which is suitable for the MH decontamination requirements. It is powered by a

 Proc. ACM Interact. Mob. Wearable Ubiquitous Technol., Vol. 5, No. 3, Article 119. Publication date: September 2021.
VeriMask: Facilitating Decontamination of N95 Masks in the COVID-19 Pandemic: Challenges, Lessons Learned, and Safeguarding the Future
119:10 • Yan Long, Alexander Curtiss, Sara Rampazzi, Josiah Hester, and Kevin Fu

 Sensor Node I2C
 LED Laird BL653 Si7021
 Button
 nRF52833
 Battery
 Node
 Button Interrupt
 N95 Mask Advertsing/Idle
 BLE
 Sensing Scheduler
 Advertising LED Indicator

 Paper Towel
 Android App User Config Display Activity
 Export Data
 Node Manager Detection&Alarm
 BLE Scanner
 Decon Verifier Optimization

Fig. 3. VeriMask Architecture. The VeriMask platform consists of BLE sensor nodes acting as peripheral data collectors, and
an Android application acting as the central monitor and verifier.

Fig. 4. (Left): The prototype VeriMask V1 sensor node attached to the lid of the Pyrex container along with the N95 mask
and the paper towel. (Middle): VeriMask V1 sensor node. (Right): VeriMask V2 sensor node.

high-temperature (125°C maximum), lithium poly-carbonmonofluoride coin cell battery with 550 mAh capacity
(Panasonic BR2450A).
 The sensor node V1 board includes sockets for two temperature and relative humidity sensors’ breakout boards,
namely the Adafruit Si7021 containing Silicon Labs’ Si7021-A20 sensor [4], and the Sensirion SHT8 containing
Sensirion’s SHT35-DIS sensor [5], to facilitate testing and replacement of the sensors. The sensor nodes also
include a user button to stop and restart the data transmission as well as an LED which blinks while collecting
and transmitting data as an extra indicator for the operators. We then label each node with its nodeID on the
back which corresponds to the nodeID in its BLE broadcasting packets, and attached the nodes on the lids of
the containers to avoid contact with the masks, and with their back side facing upward so that the operators
can easily see the nodeID through the transparent lids as shows in Figure 4. Sensor node V2 was designed with
the design goals of low cost, simplicity, reliability, and manufacturability. We greatly reduced the number of
components, brought down the board to two layers, and reduced the size by more than half, measuring only
3.5cm x 3.8cm (1.4 inches x 1.5 inches). As a result the cost was reduced from $38.27 per board (V1) to $15.66
including components, manufacture, and assembly from a U.S. based manufacturer at a quantity of 1000.
The number of sensor nodes. VeriMask uses a dense one-for-one topology by associating one sensor node to
each individual mask. The choice of this topology compared to others (e.g. one-for-many [69]) depends on two
main factors. First, using a single sensor to monitor a group of masks requires that the heat and humidity gradients
in the sensor node area should be small and should not be subjected to significant variations. However, as shown

Proc. ACM Interact. Mob. Wearable Ubiquitous Technol., Vol. 5, No. 3, Article 119. Publication date: September 2021.
Facilitating Decontamination of N95 Masks • 119:11

in previous work [45, 78, 80] and Figure 2, it is strongly dependent on the heating device characteristics and the
MH decontamination process. Thus, the number of sensors nodes used should vary based on a careful thermal and
humidity profiling of every heating device as well as every state of the decontamination procedure beforehand.
This prevents the rapid deployment of the decontamination technology and makes it impractical, burdensome,
and error-prone for non-specialized operators. The second factor is the use of separate sealed containers for
each mask. The individual containers are used in MH decontamination processes to avoid cross-contamination
between masks and shape deformation, and also to provide the adequate humidity environment for an effective
decontamination. Temperature and humidity fluctuations (e.g. leakage) in sealed containers are extremely difficult
to measure from outside the containers and might lead to undetected decontamination failures.
Sensor selection. The temperature and relative humidity sensor is a crucial component of the VeriMask platform.
The sensor should be chosen based on its reliability to accurately detect changes in the environmental conditions
during decontamination, its cost, and its power consumption. Figure 5 (a) shows the critical parameters of four
potentially suitable sensors designed to operate under high temperatures (125°C maximum) [2, 4, 5]. We consider
the sleep currents as the most important parameter for identifying power consumption of the sensors because
for MH decontamination monitoring scenarios the temperature and relative humidity change slowly over time
(and the required sample rate is extremely low), thus the sensor would be in sleep mode for most of the time.

 80
 Temp (°C)

 60
 40 Temperature Si7021
 Temperature SHT85
 20
 0 50 100 150 200
 Relative Humidity Si7021
 100 Relative Humidity SHT85
 RH (%)

 50

 0
 0 50 100 150 200
 Time (min)
 (a) (b)
Fig. 5. (a) Cost, sleep current (Sleep Crt.), relative humidity/temperature precisions (RH/Temp. Prec.) and relative humidi-
ty/temperature response time (RH/Temp. Resp.) of 4 high-temperature resistant sensors. We identify the sensor Si7021-A20
as the most suitable for our VeriMask platform. (b) The temperature and relative humidity readings of Si7021-A20 sensor and
the SHT35-DIS sensor in 5 consecutive heating cycles.

 To find the most suitable sensor for VeriMask, we conducted comparative experiments with the sensor Si7021-
A20, which has the lowest cost and power consumption, and the sensor SHT35-DIS, which has the best response
times and precision. We used the Sun Electronics EC12 thermal chamber as the convection heating device, and
set the time for each decontamination cycle ( ) to 40 minutes. We did not preheat the thermal chamber so
the first heating cycle of Figure 5 (b) shows how the sensors behave when no preheating is performed before
decontamination. We use two VeriMask sensor nodes in the chamber, each with one Adafruit Si7021 and one
Sensirion SHT85 breakout board (see Figure 4), and perform five consecutive cycles in order to examine the
long-term reliability of these sensors.
Experiment results. Figure 5 (b) illustrates the averaged readings of the two VeriMask nodes during the 5-cycle
experiment. In the first heating cycle, the maximum and average (absolute) differences in the two sensors’
temperature are 4.8°C and 1.8°C, and those in the relative humidity are 22% and 6.8%, respectively. However,

 Proc. ACM Interact. Mob. Wearable Ubiquitous Technol., Vol. 5, No. 3, Article 119. Publication date: September 2021.
119:12 • Yan Long, Alexander Curtiss, Sara Rampazzi, Josiah Hester, and Kevin Fu

during the following cycles, the numbers decrease to 1.8°C and 0.2°C for temperature differences, and 3% and
0.8% for relative humidity. The differences of readings between the two sensors are generally negligible except
for the first cycle, where Si7021-A20 measured a relative humidity overshoot at the beginning, and reported
lower temperatures than the SHT35-DIS sensor. The possible cause is that Si7021-A20 has slower temperature
response time, and the lower temperature resulted in a higher relative humidity [51]. However, the two sensors
had comparable performance in the following cycles, suggesting the Si7021-A20 sensor with slower response time
and lower precision is suitable for the MH decontamination monitoring. In addition, using the sensor Si7021-A20
for VeriMask helps to optimize for power consumption and costs. We thus decided to only use Si7021-A20 for
VeriMask sensor node V2. All experiments in the evaluation sections are conducted with Si7021-A20 only.
Considerations for decontamination. Another key finding from the above experiment is that without pre-
heating the heating system with the containers, the transient time will generally be significantly longer. This
could result in insufficient decontamination if the process time is configured to be the same as in a preheated
system. To take Figure 5 (b) as an example, while the 40 minute decontamination cycles is enough to keep the N95
masks in the expected decontamination ranges for over 30 minutes with a lower decontamination temperature
threshold ( ) of 70°C, it is not sufficient for the first cycle. The appropriate time for the first decontamination
cycle also depends on the specific heating system’s heat rate and temperature as well as the expected temperature
range. Configuring a separate decontamination period for the first cycle is thus non-trivial. As a result, we
recommend to preheat the device to the decontamination temperature before the decontamination starts. We will
further discuss the potential failures without preheating in Section 6.

4.2 BLE Communication
The sensor nodes use non-connectable BLE advertising to broadcast sensor data to the Android monitor application.
Opting for wireless sensor nodes makes the system flexible and easier to maintain by eliminating the need of
wiring for each (potentially hundreds of) sensor nodes. We choose non-connectable BLE advertising over other
active BLE modes and synchronous wireless communication methods (e.g. WiFi) for two main reasons. First, these
communication schemes limit the number of peripherals that can communicate with the central simultaneously,
and increase the power consumption of the sensor nodes as well as the time for adding new nodes into the
system [49, 85]. These factors significantly reduce the effectiveness of the platform. Second, custom wireless
protocols are generally less supported by commercial mobile devices compared to BLE broadcasting, which is the
most common protocol used in smartphones. In Section 5.2 we further demonstrate that with the information
redundancy and the low probability of critical information loss, the benefit of having active control is negligible.
BLE broadcasting implementation. The manufacturer-specific data field of the advertising payload contains
node identification numbers (nodeIDs) and sensor data. Out of the 31-byte payload space of the non-connectable
BLE advertising packet, we use 12 bytes in total with 1-byte nodeID and 4-byte sensor data. The 1-byte nodeID
field supports simultaneous decontamination verification of up to 256 N95 masks in a local sensor network. The
number can be further increased by utilizing the remaining 19-byte payload space. The sensor nodes broadcast
with a transmission power level of 0 dBm to balance power consumption and transmission success rate.
 Figure 6 (a) shows the BLE advertising scheme of VeriMask. Each node advertises with advertising interval
 = 10 by default, which is the time between two consecutive advertising events. The 10s interval is determined
empirically based on preliminary experiments’ results that temperature and RH have a extremely slow change
rate during the decontamination process, and the fact that BLE advertising supports a largest advertising interval
of 10.24s. For each advertising event, the node sends three advertising packets on three different frequency
channels (channel 37, 38, 39) [82]. Loss of packets can happen due to packet collisions (e.g. packet overlapping in
time) on the same channel between different nodes in multi-node scenarios. A 0-10 ms random delay is added
before every advertising event in order to mitigate collisions, as is specified in the BLE protocol [6]. If too many

Proc. ACM Interact. Mob. Wearable Ubiquitous Technol., Vol. 5, No. 3, Article 119. Publication date: September 2021.
Facilitating Decontamination of N95 Masks • 119:13

 Collision Period Collision Period

 Adverstising Event Adverstising Event Adverstising Event Adverstising Event

 Ch 38

 Ch 38
 Ch 39

 Ch 39
 Ch 37

 Ch 37
 Ch 38

 Ch 39
 Ch 37

 Ch 38

 Ch 39

 Ch 37
 Radom Radom Radom Radom
 ta Delay ta Delay
 ta Delay ta Delay
 Collision Node A advertising
 Scanning Event Scanning Event Scanning Event
 Ch 37 Ch 38 Adverstising Event Adverstising Event
 Ch 39

 Ch 38

 Ch 38
 Ch 39

 Ch 39
 Ch 37

 Ch 37
 ScanWindow (tw) ScanWindow (tw) ScanWindow (tw)
 Radom
 ScanInterval (ti) ScanInterval (ti) ScanInterval (ti) ta Delay ta

 Node B advertising
 (a) (b)

Fig. 6. (a) BLE advertising and scanning scheme (not up to scale) of the VeriMask platform. (b) Advertising collision defined
in the multi-node packet collision simulation. If any two nodes’ advertising events overlap, a collision happens.

nodes reside in the same local network, however, packet collisions might still happened even with the added
random delay. The relationship between packet collision rates and number of nodes in the system thus become
an important factor which can affect the VeriMask platform performance. Previous research works [82, 97] have
demonstrated the acceptable packet collision rates in dense BLE advertising applications. We further show in
Section 5.2 that our BLE advertising configurations guarantee low collision rates and reliable BLE transmission.

4.3 Android Monitoring Application
We choose smartphones as the front-end monitoring device over PCs due to their compactness and pervasiveness.
In order to monitor, verify, automate, and optimize the decontamination process of each mask, we built an
Android monitoring application prototype. The monitor application’s back-end comprises of functional blocks of
BLE Scanner, Node Manager, and Decontamination Verifier.
BLE Scanner. We implement the BLE scanner with the standard Android Bluetooth library [8]. In BLE scanning,
each scanning event lasts for the period of ScanWindow ( ) with an interval of ScanInterval ( ) between each
scanning event. should be less than or equal to , with an upper bound of 10.24s for both parameters. The
ratio of and is the scanning duty cycle. Figure 6 (a) shows the scanning scheme (not to scale). We configure
the monitor smartphones to be in the low-latency scan mode which provides the highest scanning duty cycle [9].
However, it is worth noting that the actual duty cycle also depends on the firmware implementation of different
manufactures. As a result, the model of the smartphone can also have an effect on the scanning behavior. We
evaluate the performance of BLE scanner with different models of smartphones in Section 5.2.
Node Manager. It is responsible for keeping track of the status of all nodes and the application itself. The
monitor application can be in one of the 3 states: IDLE, NODE_DISCOVER (adding nodes into the system), and
RECORDING (logging and verifying decontamination process). Based on its decontamination status, each node
can be in one of the 5 states: READY (ready for a new cycle of decontamination), DECON (decontamination in
progress), ERROR_FAILURE (e.g., temperature/RH does not reach the lower threshold after decontamination
starts, exceeds the upper threshold, or falls below the lower threshold, etc.), ERROR_LOST (e.g. the application
does not receive the node’s packets), and DONE (decontamination completed for this node). State transitions are
strictly controlled by finite state machine models as shown in Figure 7 (a).
Decontamination Verifier. The verifier takes the user configuration inputs of the target decontamination ranges
for temperature ( ℎ , ) and relative humidity ( ℎ , ) as well as the decontamination time ( ) within these
ranges, and compare them with the sensor reading data. If verification events (e.g., violations of decontamination

 Proc. ACM Interact. Mob. Wearable Ubiquitous Technol., Vol. 5, No. 3, Article 119. Publication date: September 2021.
119:14 • Yan Long, Alexander Curtiss, Sara Rampazzi, Josiah Hester, and Kevin Fu

 READY

 Decon
 Decon
 Stop Decon Decon
 Start
 Stop Stop
 Transient Time ERROR_
 Expires FAILURE

 Range Violations ERROR_
 LOSS
 DECON DONE
 Node Comm Loss

 Decon Completion

 (a) (b) (c)

Fig. 7. (a) The sensor node states and transitions in VeriMask Android monitor application. (b) The main control page which
shows the states of each individual mask inside the containers. (c) The detail page which displays the information of the
sensor node associated with the mask chosen on the main page.

ranges) are detected for a certain node, the verifier reports back to Node Manager which will then change the
state of that node. The criteria of verification events and state transitions shown in Figure 7 (a), are as follows.
1) Loss of nodes: If the application misses packets from a sensor node for more than a grace period of 2 minutes
continuously, the node state is changed to ERROR_LOST.
2) Decontamination Failure. Three main events can cause violations and a change of state to ERROR_FAILURE: i)
when temperature and/or the humidity level of the node does not reach the pre-configured ranges for a grace
period of 20 minutes continuously from the beginning of the MH process time (including the transient time
 ), ii) when temperature and/or the humidity level falls out of of the pre-configured ranges for a grace period of
2 minutes continuously after the transient time, and iii) when the node has not completed the decontamination
time at the end of the MH process time.
3) Decontamination completion: If the temperature and humidity level have been within the pre-configured
range for the decontamination time , the node state is changed to DONE.
Note that the time for the grace periods are the default values, which can be changed by the operator if needed.
User Interface. The user interface consists of three pages, namely i) a main control page and ii) a node details
page which are shown in Figure 7, and iii) a decontamination parameter configuration page. The main page lists
all the nodes added as well as their statistical information. Operators can check a profiling checkbox and enter
profiling mode before each decontamination cycle for throughput maximization (Section 4.4). It also displays a
timer count down for which will be determined by throughput-maximization algorithm. When selecting a
node, the application navigates to the node details page, where the most recent temperature and relative humidity
readings and other node information including node states are displayed. All pages are updated in real time.

4.4 Throughput Maximization
Beside providing monitoring capability, VeriMask has been designed to provide feedback information to help the
operators achieve the optimal decontamination throughput, i.e., achieving the maximum number of successfully

Proc. ACM Interact. Mob. Wearable Ubiquitous Technol., Vol. 5, No. 3, Article 119. Publication date: September 2021.
Facilitating Decontamination of N95 Masks • 119:15

decontaminated masks in a given time, based on the heating device used. To build this functionality, we develop
a throughput-maximization algorithm for the App that can be used in the profiling phase of the decontamination
to determine and suggest the optimal combination of decontamination parameters.
Problem Formulation. The general idea of throughput-maximization is that given a total continuous working
time selected by the operators, e.g., 8 hours, and the decontamination data of 1 profiling cycle, the algorithm
predicts the heating device temperature and MH process time ( ) that will generate the highest number of
successfully decontaminated masks in the 8 hours. The reason for maximizing the throughput in the total working
time instead of just one decontamination cycle can be easily illustrated with an extreme example: decontaminating
19 masks in 30 minutes per cycle is better than decontaminating 20 masks in 60 minutes. The algorithm is based
on the assumption of similar dynamics of the profiling cycle with the following decontamination cycles after 1
preheating cycle, as verified in our experiments (e.g., the last 4 cycles shown in Figure 5).
 We start by looking at the general optimization problem, where no profiling or other data is given. We denote
the selected total working time as , the time for the conducted profiling cycle as , the heating device
temperature as , the total number of successfully decontaminated masks in as , and that of each
decontamination cycle as . We then have the following equations:
 
 = 
 
 × 
 = ( , ),
where (·) is an unknown deterministic function with parameters such as the heating device model (its heat
transfer and distribution functions), container placements, etc. We can now formulate the general throughput-
maximization problem as the 2-D optimization problem:
 
 , = argmax 
 , 

 Note that it is almost impossible to get a closed-form analytical solution because of the unknown (·). With the
data from the profiling cycle and the assumption that each cycles have almost identical dynamics, however, we
can instead implement discrete algorithm to search for an optimized numerical solution in a simplified context.
Implementation. Intuitively, by utilizing the profiling cycle’s data we can convert the optimization problem
into a discrete 2-D search problem for a global maximum. Furthermore, we observe through experiments that the
temperature instead of relative humidity of the masks is the limiting factor of transient time in the normal case.
For example, the blue stars in Figure 12 that represent the time instance when a mask enter the DECON status
are determined by the temperatures. As a result, we only need to consider the temperature data of the masks in
the profiling cycle to determine if each mask has been in the decontamination range.
 Varying can be easily done on the profiling cycle’s data by fixing the start of at time 0 while
changing the cycle end time. Varying , however, is more tricky since we only have the data for 1 profiling
 (0)
cycle where the heating device temperature was fixed to a certain value . To sweep through different 
while avoiding increasing the number of required profiling cycles, we use a stretching method that adapts
the temperature data of each mask collected in the profiling cycle to temperature curve estimations under
different . Based on the observation that changes in does not change the overall trend (the type of
underlying time function) of the temperature data curves, we stretch the curves proportionally according to
 (0) 
 / . After that, we can finally apply the optimization (search) algorithm to find the best set of 
 
and . We summarize the algorithm for the above throughput-maximization in Algorithm 1. In the App
implementation, = [ : 1◦ : 100◦ ], = [ : 1 : ], where the 2nd element of each
vector represents the discrete step size. After the profiling cycle, the App runs the throughput-maximization
 
algorithm and reports the optimal decontamination parameters and as well as the expected number

 Proc. ACM Interact. Mob. Wearable Ubiquitous Technol., Vol. 5, No. 3, Article 119. Publication date: September 2021.
119:16 • Yan Long, Alexander Curtiss, Sara Rampazzi, Josiah Hester, and Kevin Fu

 Algorithm 1: Throughput Maximization
 Input: selected total working time , profiling cycle temperature data matrix , profiling cycle heating device temperature
 (0)
 , required in-range decon time , decon temperature thresholds [ , ℎ ], optimal heating device temperature
 (candidate) vector and MH process time (candidate) vector 
 
 Output: , , 
 
 1: Initialization: ← 0, ← 0, ← 0
 2: for each candidate do
 (0)
 3: ℎ = ℎ ( , , )
 4: for each candidate do
 5: = ( , ℎ , , [ , ℎ ])
 
 6: if > then
 
 7: = , = , = 
 8: else
 9: Do Nothing
10: end if
11: end for
12: end for
 
13: return , , 

of successfully decontaminated masks using these parameters. The App will also inform the operators
of the container locations where the masks are predicted to experience decontamination failures so that the
operators can avoid placing masks there. We will further show a case study for evaluating the effectiveness of
this throughput-maximization method in Section 5.3.

5 EVALUATION
In this section, we evaluate the performance of VeriMask. We examine the power consumption of the VeriMask
sensor nodes in Section 5.1, and examine the reliability of the wireless sensor platform in Section 5.2 by looking
into the BLE transmission performance in different scenarios. We conducted these evaluations with two different
heating systems, namely a Sun Electronics EC12 thermal chamber in a laboratory setting, and a Memmert HCP
humidity chamber in clinical setting. Finally, we analyze the effectiveness of the throughput-maximization
functionality in Section 5.3 by carrying out a case study with the EC12 thermal chamber.

5.1 Power Consumption & Battery Life
Circuits’ power consumption under higher temperatures increases because of increased leakage current [56].
Due to extreme temperature variation and the high-temperature nature of the decontamination process, we
investigate the current consumption and battery life of our VeriMask sensor nodes under different temperatures.
Experimental Setup. The sensor node runs the same program as in the decontamination settings, where the
board fires one advertising event every 10s, and one of the LED flashes every 5s for 5ms as an indication of normal
operation. The EC12 thermal chamber is used as the heating device. We used an STM32 Power Shield [3] to
measure the currents, and configured the supply voltage, sample rate, sample period, minimal current threshold
to 3V (same as the supply voltage of the battery), 20000 samples/s, 100s, and 1uA respectively. We left the STM32
Power Shield outside the thermal chamber and used a long jumper wire to connect it to the sensor node to avoid
measurement errors of the STM32 Power Shield when undergoing high temperatures. When calculating the
estimated battery life, we revise down the total usable battery capacity to a conservative 70% of the specified
capacity to account for manufacturing tolerances and other conditions that can degrade battery capacity [52].

Proc. ACM Interact. Mob. Wearable Ubiquitous Technol., Vol. 5, No. 3, Article 119. Publication date: September 2021.
You can also read